Smart structures and materials: Vibration and control

Benjeddou, Ayech; Mechbal, Nazih; Deü, Jean-François

doi:10.1177/1077546320923279

Cited by 2 publications

(1 citation statement)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The CM relies on Reinforcement Learning (RL)-based methods that use iterative algorithms to converge in an optimal navigation policy [30][31][32]. It is common in MAS to use methods based on Deep Reinforcement Learning (DRL), which is a powerful tool that combines neural networks and RL algorithms that allow each agent to learn from its interactions with the environment [33][34][35][36][37][38]. Despite the effectiveness of the RL-based methods, the main disadvantage in MAS is the computational complexity and abundance of data required to converge to the global policy.…”

Section: Introductionmentioning

confidence: 99%

Decentralized Navigation with Optimality for Multiple Holonomic Agents in Simply Connected Workspaces

Kotsinis,

Bechlioulis

2024

Sensors

View full text Add to dashboard Cite

Multi-agent systems are utilized more often in the research community and industry, as they can complete tasks faster and more efficiently than single-agent systems. Therefore, in this paper, we are going to present an optimal approach to the multi-agent navigation problem in simply connected workspaces. The task involves each agent reaching its destination starting from an initial position and following an optimal collision-free trajectory. To achieve this, we design a decentralized control protocol, defined by a navigation function, where each agent is equipped with a navigation controller that resolves imminent safety conflicts with the others, as well as the workspace boundary, without requesting knowledge about the goal position of the other agents. Our approach is rendered sub-optimal, since each agent owns a predetermined optimal policy calculated by a novel off-policy iterative method. We use this method because the computational complexity of learning-based methods needed to calculate the global optimal solution becomes unrealistic as the number of agents increases. To achieve our goal, we examine how much the yielded sub-optimal trajectory deviates from the optimal one and how much time the multi-agent system needs to accomplish its task as we increase the number of agents. Finally, we compare our method results with a discrete centralized policy method, also known as a Multi-Agent Poli-RRT* algorithm, to demonstrate the validity of our method when it is attached to other research algorithms.

show abstract